One hundred and nine gastric cancer patients (males 41, females 68, age range 32–89 years) (25 cases Dukes’ A, 22 cases Dukes’ B, 28 cases Dukes’ C and 34 cases of Dukes’ D), and 106 cases of controls (males 50, females 56, age range 26–85 years) (60 healthy volunteers, 16 cases of chronic atrophic gastritis patients, and 30 cases of chronic superficial gastritis) were recruited for this study at the First Affiliated Hospital of Zhejiang University and Zhejiang Taizhou Municipal Hospital. The subjects were assigned into an experimental group and a verification group according to Table I. The peripheral blood samples were collected in the morning after overnight fasting. The blood samples were then maintained at 4°C for 1–2 h prior to centrifugation at 3,000 rpm at 4°C for 10 min to separate out the serum. The serum samples were frozen at −80°C in a freezer for storage. All protocols and experiments were approved by the Taizhou Medical College Ethics Committee for clinical experiments and use of human samples; written informed consent was obtained from all subjects participating in this study. The study complied with the World Medical Association Declaration of Helsinki regarding ethical conduct of research involving human subjects.

After thawing and 10 min of centrifugation (10,000 rpm), a 20-μl serum sample without fraction treatment (which does not affect the protein mining efficiency as shown by the authors) was added to 30 μl 0.5% U9 (9 mol/l urea, 2% CHAPS (3[(3-cholamidopropyl)dimethylammonio]-l-propanesulfonate), 1% DTT (DL-dithiothreitol)) in a 96-well plate and incubated for 20 min at 4°C with 600 rpm vigorous agitation. The ProteinChip array cassette was put into a 96-well bioprocessor and 100 μl U1 buffer (50 mmol/l Tris-HCL diluted 10% U9 buffer) was added into each well, and incubated for 10 min at 4°C with 600 rpm vigorous agitation. Q10 buffer (200 μl) (100 mM Tris-HCl buffer pH 9.0) was then added and a 5-min incubation was carried out 2 times with agitation. All experimental reagents were obtained from Shanghai Shenggong Company, Shanghai, China.

Fifty microliters of the protein-denatured serum samples were removed to a new tube, and 200 μl Q10 buffer was added to dilute the samples before being applied onto the Q10 chip Bioprocessor (Ciphergen) for 60 min. Then each plate of the Q10 chip was added together with 200 μl Q10 buffer, incubation was carried out for 5 min two times with agitation, and finally 20 mm/l HEPES (pH 7.4) buffer was added for washing before drying. SPA (0.5 μl) was added 2 times into each plate with drying between each addition. Mass spectrometry was set as laser intensity 185, sensitivity 8, 2,000–20,000 m/z. Inter-chip CV was <10%. All-in-one control chip was used to adjust the system with a systemic error <0.1%. All of the data were processed with ProteinChip 3.0 software, then the Biomarker Wizard software 3.1 and Biomarker Wizard software 4.0.1. P<0.01 was determined to indicate statistically significant differences in the comparison between two protein peaks.

Serum samples (100 μl) with 300 μl water and 700 μl acetonitrile were mixed and maintained at −20°C in a freezer for 30 min prior to centrifugation at 3000 rpm for 10 min. The supernatant was freeze-dried for 20 min before collection for HPLC. The purified solutions were collected at different time periods, freeze-dried to obtain a 20-μl volume solution. Solution (0.5 μl) was mixed with 1.5 μl matrix solution (10 mg/ ml CHCA) to be placed on chip points, with crystallization for MALDI-TOF MS detection. The conditions included pulsed nitrogen laser (337 nm), accelerating voltage 20 KV, linear analysis mode, mass range 3,000–20,000 m/z. The samples corresponding to the specific protein peaks in SELDI-TOF MS were subsequently identified.

Purified target protein of 20 μl was mixed with 60 μl 8M urea (final concentration of urea 6 M) and was agitated at room temperature for 20 min. Then 0.8 μl 1 M DTT (final concentration 10 mM) was added and mixed at room temperature for 1 h, and 3.2 μl 1 M iodine acetyl amine (final concentration 40 mM) was added and maintained for 45 min in the dark. DTT (3.2 μl 1 M) (final concentration 40 mM) was added for 20 min, and then 400 μl 50 mM NH₄HCO₃ was added to dilute the solution, with urea concentration at 1 M and pH 8.0. Subsequently, 0.1 μg protease in a 37°C water bath for 1 h, and formic acid was adjusted to a pH <3 to terminate the reaction. The hydrolysates were subjected to LC-MS/MS analysis. Sample solutions were put in a self-made C18 capillary column for liquid chromatography: inner diameter 100 μm, filled part 100 mm, filled particles with diameter 5 μm. The flow phase A was water and 0.1% formic acid; the flow phase B was acetonitrile and 0.1% formic acid. The washout followed the sequences below: 100% A (0 min) - 100% A (5 min) - 5% B (5.1 min) - 65% B (60 min) - 100% B (75 min) - 100% B (85 min). The flow speed was 200–800 nl/min. The data-dependent mode was used; the scanning ranges were from 400 to 2,000 m/z; the five strongest signal peaks of each full scan were selected for secondary MS (MS2) analysis.

The data retrieval used the XCalibur program components BioWorks 3.2 (Thermo Finnigan) and the peptide sequences were searched for in the NCBI human protein database (human.ref) according to the mass spectrum. The following parameters were used: enzyme split site at random site; fixed modification at cysteine amine formylation modification; variable modification at methionine oxidation; retrieval parameters ΔCN >0.1; Sp >500; Rsp ≥5; Xcorr vs. Charge: Xcorr (+1) >1.9, Xcorr (+2)>2.5, Xcorr (+3) >3.75.

The protein fingerprint spectrum from 65 cases of gastric adenocarcinoma and 53 cases of control were normalized first, then analyzed using Biomarker wizard software. Two hundred and twenty-seven protein peaks were found in the m/z range from 2,000–50,000 (Fig. 1).

The Biomarker Wizard software analysis revealed three differentially expressed protein peaks in serum samples from gastric adenocarcinoma patients, with m/z at 5,906.5, 6,635.7 and 8,716.3 respectively (P<0.01) (Fig. 2 and Table II), which were considered as potential biomarkers for gastric adenocarcinoma. The 5,906.5 peak showed increased expression in the gastric adenocarcinoma cases; while the 6,635.7 and 8,716.3 peaks showed a decreased expression level in the gastric adenocarcinoma serum samples (Table II). The combined analysis with the three peaks as the basis for the diagnostic model showed a sensitivity of 93.85 (61/65) and a specificity of 94.34% (50/53) in analyzing the mass spectrometry data from the 65 gastric adenocarcinoma patients and 53 cases of control subjects (Table III).

The protein peaks of the three biomarkers (m/z 5,906.5, 6,635.7 and 8,716.3) were isolated and purified with HPLC, and then collected into PCR tubes for MALDI-TOF-MS examination. The results showed that the three peaks were proteins with molecular weights of 5,906.4, 6,632.9 and 8,704.3, respectively (Fig. 3).

Following LC-MS/MS measurement of the digested proteins and the data screening from NCBI human protein database, the 5,906.4 peak was found to correspond to the fibrinogen α chain (100% match), the 6,632.9 peak corresponded to the apolipoprotein A-II (96% match), and the 8,704.3 peak corresponded to the lipid-laden protein C-I (60% match) (Table IV).